Meta reaches for its own silicon
The day belonged to Meta. According to an internal memo cited by Reuters, the company is preparing to build its own artificial-intelligence chips, with production of a homegrown processor codenamed Iris slated to begin in September. It is a move that could loosen one of the tech industry's tightest dependencies: its reliance on Nvidia for the hardware that powers modern AI.
Wall Street reacted quickly. Meta shares jumped nearly 7.5% to $677.90, capping the stock's best week since 2024. After nearly a year in the doldrums, traders are falling back in love with the company — CNBC framed its AI push as starting to look like "the recipe" investors had been waiting for. If a customer as large as Meta can supply even part of its own compute, the ripple effects reach every chipmaker that has ridden the AI wave, and Nvidia most of all.
The memory boom gets its own gold rush
If logic chips grabbed the headlines, memory chips supplied the week's biggest structural story. SK Hynix landed on Wall Street in a trillion-dollar debut, a marker of how far the scramble to feed AI has pushed a company once seen as a commodity supplier. The same firm is now signaling how long the squeeze may last: SK Hynix says shortages of the specialized memory that AI systems depend on could peak in 2027 — meaning the crunch that has defined the market may get worse before it eases.
That outlook helps explain the capital now pouring into production. Micron has raised its planned U.S. investment to a staggering $250 billion, aimed squarely at expanding domestic manufacturing of the memory chips that AI runs on. Between SK Hynix's public-market splash and Micron's factory bet, the message is consistent: demand is being treated as a durable, multi-year phenomenon, not a spike.
SK Hynix is also hedging toward efficiency. The company has teamed with startup TetraMem and the University of Southern California on an experimental memristor chip designed to cut the power AI consumes on "edge" devices — the phones, sensors, cameras and gadgets where energy budgets are tight. It is a reminder that the race isn't only about more capacity, but about doing more with less.
Japan's price play against TSMC
Manufacturing leadership drew a new challenger. Japan's Rapidus is trying to break into the most advanced corner of the business — 2nm-class silicon — by competing on price rather than prestige. According to disclosures reported by Tom's Hardware, the company plans to undercut Taiwan's TSMC on its cutting-edge wafers, a bold pitch given how few firms can produce chips at that node at all. Whether customers will trust a newcomer with their most valuable designs is the open question, but the strategy signals that even the frontier of chipmaking is no longer a one-horse race.
And for the gamers
Amid the industrial-scale deals, a consumer note: NVIDIA is rumored to be preparing a GeForce RTX 5090 SE, a card that would slot between the RTX 5080 and the flagship RTX 5090. Per TweakTown, it is positioned to deliver near-flagship performance for roughly $500 less — a tempting middle ground for enthusiasts priced out of the top tier.
The through-line
Taken together, the day sketches an industry in motion on every front at once. The biggest buyers, like Meta, are becoming makers. The memory suppliers are cashing in and warning the shortage has years to run. Governments and companies are committing hundreds of billions to new factories, and even the most exclusive club in manufacturing is facing a price-based challenger. Nvidia still sits at the center of it all — but for the first time in a while, the gravity around it is shifting.